• Aucun résultat trouvé

Postulation and gonality for projective curves

N/A
N/A
Protected

Academic year: 2022

Partager "Postulation and gonality for projective curves"

Copied!
12
0
0

Texte intégral

(1)

R ENDICONTI

del

S EMINARIO M ATEMATICO

della

U NIVERSITÀ DI P ADOVA

E DOARDO B ALLICO

Postulation and gonality for projective curves

Rendiconti del Seminario Matematico della Università di Padova, tome 80 (1988), p. 127-137

<http://www.numdam.org/item?id=RSMUP_1988__80__127_0>

© Rendiconti del Seminario Matematico della Università di Padova, 1988, tous droits réservés.

L’accès aux archives de la revue « Rendiconti del Seminario Matematico della Università di Padova » (http://rendiconti.math.unipd.it/) implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/conditions).

Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte- nir la présente mention de copyright.

Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques

http://www.numdam.org/

(2)

Gonality Projective

EDOARDO BALLICO

(*)

We are interested in the

interplay

between intrinsic and

projective properties

of curves. In

particular

we are interested in the

postulation

of

general k-gonal

curves. A smooth curve Y c PN is said to be canonical

if

1".1 Ky.

A curve Z c PN is said to have maximal rank if the restrictions maps

r,,,(k): HO(PN,

-

HO(Z, c~~(k)~

have maximal

rank for all

integers

k. In

[4]

it was

proved

that for every N &#x3E;

3, g &#x3E; N,

a

general non-degenerate

canonical curve with genus g in PN has maximal rank.

Here we prove the

following

results

(over C).

THEOREM 1. For all

integers

N &#x3E;

3, g

&#x3E;

N,

a

general trigonal (resp. bielliptic) non-degenerate

canonical curve of genus g in PN has maximal rank.

THEOREM 2. For all

integers N, d,

g,

with g

&#x3E; N &#x3E;

3,

d &#x3E;

2g,

a

general embedding

of

degree d

in PN of a

general hyperelliptic

curve

of genus g has maximal rank.

The

proofs

of theorem 1 and theorem 2 is modulo a

smoothing

result

given in § 1,

almost the same that the

proofs

in

[4];

in

parti-

cular

in §

3 we omit the details which can be found in

[4]

or

[3].

For

§

1 we use the

theory

of admissible

coverings ([7])

which is very useful to obtain results about

general k-gonal

curves

by degeneration

tech-

niques.

The inductive method

of § 2, § 3, (the

Horace’s

method)

was

introduced in

[8].

(*) Indirizzo dell’A.:

Dipartimento

di Matematica, Università di Trento, 38050 Povo

(TN), Italy.

(3)

In §

4 we show

that, by

the

theory

of admissible covers, the results

proved

in

[6]

about the minimal free resolution of

general

canonical

curves are true

(with

the same

proof)

for

general k-gonal

curves for

suitable k.

0. Notations.

Let V be a

variety (over C)

and S a closed subscheme of

V; 3s,v

is the ideal sheaf of S in V and

Ns,v

its normal sheaf

(or

normal

bundle).

Assume that we have fixed an

embedding

of V into

Pk,

so that

0v(t)

and are defined. Then

rs,v(t): HO(V,

--~ is the

restriction map. If V =

Pk,

we write often

rs,k(t), Ns,k

instead

of

rs,v(t), IS,V .

0

A curve T c PNis called a bamboo of

degree

d if it is

reduced,

con-

nected,

with at most nodes as

singularities, deg (T)

=

d,

its irre-

ducible

components

are

lines,

and each line in T intersects at most two other irreducible

components

of

T; equivalently,

we may order the lines

.L1,

... ,

Z~

of T so that iff and

only

if

A

connected,

reduced curve X c

PN, deg (X)

=

2d,

..g with

only

or-

dinary

nodes as

singularities,

is called a chain of d conics if its irre- dicible

components C1,

... ,

C,

are

conics, 0

if and

only

if

I i - j 2,

card

(Ci

n

Ci+,)

= 2 if 0 i d. Sometimes we will

allow the

reducibility

of some of the conics

Ci

i in a chain of

conics ;

if

Ci

is

reducible,

we assume that every line of

Ci

intersects the

adjacent

conics. A line

(resp.

a

conic)

of a bamboo

(resp.

chain of

conics)

T

is called final if it intersects at most another irreducible compo- nent of T.

We will write

( (a ; b))

for the binomial

coefficient ;

thus

( (a ; b )) : = : _ (a ! ) / ((a - b ) ! b ! ) .

A

triple

of

integers (d, g ; N)

with

d &#x3E; g + N,

N &#x3E;

2, g ~ 0

has critical value k is the first

integer

t &#x3E; 0 such that td

+

1- gc

((N + t; N)).

We define

integers r(k,

g,

N), q(k,

g,

N) by

the

following

relations:

A smooth curve T of

degree d

and genus g in PNwith

r(k - 1,

g,

N)

and

hl(T,

2

k,

has critical value 1~

(i.e.

(d, g; N)

has critical value

1~) ;

T has maximal rank if and

only

if

-

1 )

is

injective

and

r,,,(k)

is

surjective (Castelnuovo-Mumford’s

Lemma).

(4)

We define

integers c(k, N), e(k, N) by

the

following

relations :

A chain F of

c(k, N)

conics in PN has critical value

1~;

it has maximal

rank if and

only

if

rF,N(k

-

1)

is

injective

and

HO(PN,

==

e(k, N).

Define

integers y(k, N), k

&#x3E;

0,

N &#x3E;

2,

in the

following

way. Set

y(1, N) := c(l, N)

and assume defined

y(k - 1, N).

Set

y(k, N) _ e(l~ - 1, N), e

= 0 otherwise. Hence

N) &#x3E; y(k, N)

&#x3E;

N) -

k.

Define

integers N), j(k, N) by

the

following

relations :

A canonical curve C of

degree

d in PN has critical value k &#x3E; 1 if and

only

if

2x(k - 1, N)+2d2x).

Set

y’ (k, N) : = y(k, N) - - [(N + 5)/3]-1.

A finite subset S c Pk is said to be in Linear

general position

if

every subset W of S spans a linear space of dimension min

(k,

card

(W) - 1).

..

A

bielliptic

curve is a

smooth, connected, complete

curve with a

degree

two

morphism

onto an

elliptic

curve.

Let Y = A u B c

P3,

7 A chain of

conics, B bamboo, A

intersect-

ing

B at a

unique point, P,

and

quasi-transversally,

P

belonging

to

a final line of B and a final conic of A. An irreducible

component

of Y

is called free if it intersects

only

another irreducible

component

of Y.

1. A

smooth,

connected curve Ec Pn is called canonical if Let

C(g, n)

be the closure in Hilb

(pn)

of the set of

smooth canonical curves of genus g in

Pn ,

and

C(g,

n,

k) (resp. C(g,

n;

biel))

the closure in Hilb of the set of smooth canonical curves

of genus g in

Pn,

which are

k-gonal (resp. bielliptic)

as abstract curves.

PROPOSITION 1.1. Fix a smooth canonical

trigonal

curve C c

pn, deg (C)

=

2g - 2,

2

points A,

B in the same fiber of on

C,

and a

chain D of r conics in

Pn,

7 D

intersecting

C

quasi-transversally

and

exactly

at A and

B,

A and B

belonging

to a final conic of D. Then

C u D c- C(g + r, n, 3).

PROOF.

By

definition

C(g +

r, n,

3)

is closed in Hilb

(Pn).

Hence

we may assume

C, A,

D

general.

We know that C u D E

C(g, n)

([5], [4], § 2~ ~

(5)

First assume r = 1.

Taking

a

projection,

we may assume n = g, C V D

spanning P",

C

spanning

a

hyperplane M,

and

hl(C, NO,M)

= 0.

We know that

hl(C

U

D,

0

([4], proof

of

2.1),

hence C u D

is a smooth

point

of Hilb

(Pn).

Since C u D is

semi-stable,

we have a

morphism h

from a

neighborhood

of C u D in Hilb

(P")

to the moduli scheme of stable curves of genus

g + 1

such that

h(C

u

D)

is

the curve C’ obtained from C

pinching together

the

points ~,

B.

By

the

generality of C, A,

we may assume Aut

( C’ )

_

{1},

i.e. C’ is a

smooth

point

of

Mg+1.

To obtain 1.1 for r =

1,

it is sufficient to check that h is

flat,

hence open, at C U D.

By

the smoothness of Hilb

( lPn)

and

Mug+1

at the

corresponding points,

it is sufficient to check that the fiber has the

right

dimension

n2 +

2n = dim

(Aut ( l~n))

in a

neiborhood of C u D. A

priori

near C U

Dh-’(C’)

contains either curves

abstractly isomorphic

to C’

(i.e.

irreducible canonical stable

curves)

or curves

isomorphic

to C U D. The first

type

of curves has dimension

n2 -i--

2n. Since Pic

(C

u

D)

has a 1-dimensional

non-compact factor,

we see

easily that,

up to

projective transformations,

there is

exactly

a one dimensional

family

of curves C" U C V D. However for any 2

triples

=

1, 2, 3,

of distinct

points

of

D",

there is

with

Fi

for every i. Hence the stabilizer of any C" U D" in Aut is

one-dimensional, concluding

the

proof

of

the case r = 1.

By

induction on r, if r &#x3E; 1 it is sufficient to prove the

following

claim

stronger

than the case r =

1 just

proven.

Claim. Assume r = 1 and fix 2

general points

of

D;

then

there is a flat

family X - T,

T smooth irreducible affine curve, Xc

T X P",

with =

.~t smooth,

canonical and

trigonal

for

t E

T,

t ~

0,

and a

family

mt of

3-coverings,

mt:

Xi -

PI such that is the

given g3 , molD

sends

A, B

to one

point

of Pi and

E, F

to another

point

of l~l.

By

the

theory

of admissible

coverings ([7])

there is a

morphism

b : with

b(O)

=

C’, b(t)

a smooth

3-gonal

curve for 0.

By

the first

part

of the

proof

we may assume

Xt = b(t). By [7], proof

of we the existence of a

family

’int: -

t E

which,

as t goes to

0,

tends to an admissible

covering

mo

with =

mo (B ),

=

Taking

a suitable fiber

product,

we obtain the claim.

PROPOSITION 1.2. Fix a canonical

bielliptic

curve C c

pn, deg (C)

_

- 2g - 2,

2

points A,

B on C with the same

image

under the 2 to 1

map of C to an

elliptic

curve, and a chain D of r conics in D inter-

(6)

secting

C

quasi-transversally

and

exactly

at A and

B,

A and B

belong- ing

to a final conic of D. Then C U D E

C(g +

r, n;

biel).

PROOF. The

proof

of 1.1 works with two minor twists. We use

the notations introduced in the

proof

of 1.1. Since Aut

(C)

and

Aut

( C’ )

is not

trivial,

is not smooth at C’. Instead of

we may however use

(over C)

the Kuranishi local deformation space

ov C’

(or

a suitable

rigidification

of Instead of admissible cover-

ings

of

~1,

we have to use admissible

2-coverings

of curves of arithmetic genus 1. Since these

coverings

are

cyclic,

there is no need here of a

general theory.

Take a

2-covering

c:

C --~ Z,

Z

elliptic

curve, with

c(A)

=

c(B) (hence

c not ramified at

A, B )

and a

2-covering

d :

with

d(A)

=

d(B) (hence

unramified at A and

B).

Take as Z U ]p&#x3E;1 the

glueing

of Z and P’

along c((A)

and

d(.E).

Then c, d induce a 2-cover-

ing u :

C U D - Z U PB Let ... , a2g be the ramification

points of u,

with a, E Z if and

only

if

i o 2g -

2. Take any flat

family

s : W -

T,

with

Wo

= Z U

P’7 W,

smooth

elliptic

for t ~

0, and,

in an etale

neighborhood

of

0,

any

2g

sections sl, ..., s2g of s with

8,(0)

= ai:

The divisor

-~-

...

+ s29 (t )

on

Wt

induces a

cyclic 2-covering

which

tends to u when t goes to 0. 0

Let

Z(d, g,

n,

k) (resp. Z(d, g,

n;

biel))

be the closure in Hilb of the set of

smooth, connected, k-gonal (resp. bielliptic)

curves of

degree d,

genus g, and with non

special hyperplane

section.

Z(d,

g, n,

k)

and

Z(d,

g, n;

biel)

are irreducible

(and

not

empty

if

d ~ g + n).

We

have the

following

result.

PROPOSITION 1.3. Fix C c

Pn,

C E

Z(d, g,

n,

k) (resp. Z(d, g,

n;

biel)),

2 smooth

points A,

B on C with the same

image

under an admissible

covering

of

degree k (resp.

a 2-to-1

covering

of a curve of genus

1)

of

C,

and a chain D of r conics in

Pn,

D

intersecting

C

quasi-trans- versally,

and

exactly

at

A, B,

A and B

belonging

to the same final

conic of D. Then C U D E

Z(d + 2r, g +

r, n,

k) (resp. Z(d + 2r, g + r,

conic of D. Then C u D E

Z(d + 2r, g +

r, n,

k) (resp. Z(d + 2r,

g +

r, n ;

biel)) .

PROOF. It is much easier than the

proof

of

1.1,

1.2. Take any

flat

family

of admissible

covering, (W - T,

U -

T,

W’ --~

U)

with

Wo

= C U

D, Uo

= Z U

Pi,

= 0

(resp. 1)

and d

+

2r sections ~,vi of T with

+ ... + Wà+2r(0)

an

hyperplane

section of

Wo.

Embedd

W, using + ... + and,

if necessary,

project

the re-

sult in

(7)

We shall use often the

following

fact

([9], [2], 3.5).

Fix a smooth

curve C c 7

deg ( C)

=

d,

and a

smooth,

connected curve

D, deg (D)

= r, D

intersecting quasi-transversally

C and

exactly

at a

point.

Assume D rational and that the

hyperplane

section of C is

non-special.

Then C U D is a limit of

embeddings

of

degree

d

+ r

of C into P".

2. In this section we construct curves Y = Z U T c

P3 ,

with

Z n T =

0,

Z chain of

conics,

T

bamboo,

and with

good postulation.

This construction will be used in the next section to prove theo-

rems

1,

2 in P4.

LEMMA 2.1. Fix

non-negative integers a, b, r, s,

and a smooth

quadric Q

in P3 . Assume either

(i) a

= b &#x3E;

2,

or

(ii) a

&#x3E; b &#x3E; 1. Then there is

( Y, Z, T)

with

Z r1 T = ~, Z

chain of r

conics,

T bamboo of

degree s,

dim

( Y

r1

Q)==0y

and

PROOF. - From now on, we assume s =

0,

the

general

case

being

similar

(or

use

[1], 6.2).

If 8 =

0,

it is sufficient to prove 2.1 when

(a + 1 ) (b + 1) -

4 4r

(a + 1 ) (b + 1) +

4.

By

the properness of Hilb

(P3),

for any u and any

plane

.g there is a scheme

W,

with

Wred C H,

with

only ordinary

double

points,

W reduce outside

the

singular

locus of W limit of a

family

of chains of u conics. Just to fix the

notations,

we assume a

-~- b -~-

1 = 3 mod

(4),

the

remaining

cases

being

similar. Fix 3

general planes M, N,

1~.

Take limits

W, X, D, respectively

of chains of

[(a +

b

+ 1)/4]7 [(a -f-

b -

1)/4],

and

2, conics,

with

.M~, X, Dred c R,

W D

intersecting transversally Q,

W W X ~J D limit of a

family

of chains of

conics,

card

(D

n

Q

n

M)

=

2,

card

(D nQnN)

=2.

card

(X

n

Q

n

ll ) =1.

Set

Any

forms of

type (a, b) vanishing

on

A,

vanishes on ac

+

b

+

1

points

on

M’,

hence on ~1’.

Any

form of

type (a - 1, b - 1) vanishing

on the

points

of

n vanishes on a

+

b - 1

points

of

N’,

hence on N’. We

reduce to an assertion about forms of

type (a - 2, b - 2) (in

which

we have to consider also the 4

points

in D r1

(QB(.M’ ~J N’))~.

We

(8)

continue in the same way, never

adding

curves

intersecting

M’ U

N’ ;

for the next

step

we take .R as first

working plane.

At the end in

case

(i)

we reduce to the case a = 3 or 4

(plus

a few

points),

in case

(ii)

to cases with a 4. Consider for instance the case a = b = 3 and no

point

left from the

previous

construction

(the

worst

case).

Take 3

general

smooth curves

L, L’,

L" of

type (1, 1 )

on

Q,

and let

H, H’,

H"

the

planes they

span. The

following

chain of

conics solves our

problem.

A

(resp. A’ )

is a

sufficiently general

conic

in g

(resp..g’ ),

A" is a

sufficiently general

conic in .~"

containing

a

point

of

L,

B is a conic

containing

2

points

of L and a

point

of

L’,

but not

intersecting

L".

The aim of this section is the

proof

of the

following

lemma.

LEMMA 2.2. Fix

integers n, x,

with

n &#x3E; 29, 0 c a c 2~~ - 2.

There

exists

( Y, Z, T)

with chain of

conics,

T bamboo of

degree a, ry,3(n) surjective, deg ( Y) ~ r(n, 0, 3) -

9 - n/6

PROOF. Let 8 be the maximal

integer

with s = n mod

(4),

say n = s

+ 4t, r(s, 0, 3) - 3 (r(n, 0, 3) - a)/2 - (n - s)13. By [4], 3.1,

there is a bamboo E E

P3, deg (.E) = r(s, 0, 3) - 2,

with

surjective.

If a =

0,

set

I’ : _ .~,

T = 0. If a &#x3E;

0,

take

F c E, deg (F) = deg (E) - ~ ,

y F union of two

disjoint bamboos A, T, deg (T) _

~c. Then we fix a smooth

quadric Q

and we

apply

the so

called Horace’s method

(introduced

in

[8])

2t times. At the odd

(resp. even) steps

we add in

Q

lines of

type (1, 0 ) (resp. (0, 1)).

We

order the lines

Ll’’’.’ Ldeg(E)

of E in such a way that

Li

r1 if and

only - j

C 2. Just to fix the notations we assume s =

61~,

7~

integer (which, together

with s =

6k ~ 5,

is the worst

case).

We

add in. Q

the union U of

4k+

2 =

r(6k -f- 2, 0, 3)

-

(6k -E- 2, 0, 3) -

1

lines

A i , I i 4k + 2,

of

type (1, 0 )

with

A

i

intersecting L2i-l

for

every i. We claim that

--~- 2)

is

surj ective

f or

general

F.

To prove the

claim,

it is sufficient to find S C

P3,

(9)

in

P31Q, S"

r) S’ = f~ S"

general

in

Q.

Fix

f

E

+ 2)).

By

2.1 and the

generality

of

~",

for

general

~’ we may assume

flQ

= 0. Thus

f

is divided

by

the

equation 9

of

Q.

Since vanishes

on I’ we have =

0,

hence the claim is

proved.

Then we deform

the lines

A

to hues

A

with the

following

rule. Let

U’

be the union of the lines We assume that

U’

intersects

transversally Q. Ai

in-

tersects

Lj

if and

only

if

Ai

intersects

Li . Ai intersects Q

at a

point

on a line

B1

of

type (0, 1) intersecting L2 . Inductively,

we

impose

that

Q, j

&#x3E;

1,

has a

point

on a line

Bi

of

type (0, 1) intersecting L2

a

and a

point

on 4he line

intersecting L2i-2

and

A’-I.

The lines

Bi

are called

« good»

secants to .F’ U U’. Then we

repeat

the Horace’s method in the

following

way. To F u U’ we add in

Q4k +

4 =

=

r(6k + 4, 0, ~) - r(6k + 2, 0, 3)

lines

Bi

of

type (0, 1),

among them the

« good »

secants

previously constructed, B6k+S

linked to

L1, B6k+4

linked to Then we continue

(after smoothing

the reducible conics

obtained,

if you

prefer).

In the

step

from m - 2 to m we add

to a curve

X (m - 2)

in

Qr(m, 0, 3 ) - r(m - 2, 0, 3)

lines if

q(m, 0, 3)

&#x3E;

~ q(m - 2, 0, 3), r(m, 0, 3) - r(m - 2, 0, 3) -

1 lines if

q(m, 0, 3) q(m - 2, 0, 3) (i.e.

if m - 0 or 5 mod

(6)).

In the odd

step

from

m - 2 to m we add a certain

number,

say x, of lines of

type (1, 0),

and creates x

« good»

secants. In the even

step

from m to m

+

2

we add to

X (m)

the x

« good»

secants creates in the

previous step

and one or two lines

(always

of

type (0,1 ))

linked either to a free conic of or to a free line in a bamboo of

X(m).

We use that after 6

steps

the lines added in the 3 even

steps

are

exactly

4 more than the

lines added in the 3 odd

steps.

This

explain

the term c -

(n - s ) /3 »

in the choice of s. ·

Consider the

following

assertion

T(n,

a,

b),

defined for all

integers

n,

a, b. T(n,

a,

b ) :

There is

( Y, Z, T )

with

Z chain of a

conics,

T bamboo of

degree b,

and with

surjective.

In the

following section,

for the

proof

of theorems

1,

2 we will

need the assertion

T(n,

cc,

b)

for the values of n, a,

b,

listed in 2.3.

LEMMA 2.3. - The assertion

T(n,

a,

b)

is true if

(n,

a,

b)

has one of

the

following

values:

(2, 1, 0), (2, 0, 2 ), (3, 2, 0 ), (3, 2, 1), (4, 3, 0 ),

(4, 2, 2 ), (5, 4, 0), (5, 3, 2 ), (6, 4, 2 ), (6, 3, 4), (7, 2, 8), (7, 5, 3 ), (8, 3, 10 ),

(8, 5, 5), (9, 9, 2 ),

2

(9, 7, 4),

2

(10, 8, 5 ), (10, ~, 12 ), (11, 9, 7), (11, 5, 14),

(12, 11, 6), (12, 7, 15), (13, 14, 4), (14, 14, 8), (1~, 16, 9), (16, 18, 9),

(17, 22, 6), (18, 16, 24), (18, 22, 11), (19, 18, 24), (19, 25, 11), (20, 27, 12),

(21, 20, 33), (21, 32, 8), (22, 33, 12), (22, 25, 29), (23, 35,15), (23, 27, 30),

(10)

(24, 29, 13), (24, 30, 32), (25, 45, 8), (26, 45, 15), (27, 48, 17), (28, 52, 16), (29, 59, 10), (30, 48, 41).

Sketch of

proof.

The cases with n 24 can be done

using

several

times the Horace’s costruction

applied

not to

quadrics

but to

planes:

it is

easier;

if n

10,

we do not need any

nilpotent,

if n &#x3E; 9 we use

nilpotents

as in

[8]; only

the cases with n &#x3E; 21 are more

difficult

y however

they

can be handle also

using quadrics

as in the

proof

of 2.2.

If n &#x3E;

23,

the

proof

of 2.2 works

verbatim,

and

gives

indeed

stronger results;

for

(24, 39,13 )

start

taking

in the

proof

of 2.2 s =

12;

for the

remaining (n,

a,

b)

start from s = n - 8. ·

3. In this section we show how to

modify

the

proofs

in

[4],

to

prove theorems

1,

2. The

proof

of the case « P4 »

given

in

[4], § 8,

cannot be

adapted,

but the results proven here

in §

2 are sufficient

to prove this case and the inductive assertions of

[4]

needed for the

proofs

in

PN,

N &#x3E; 5. We will use the numbers

y(k, N),

... , intro-

duced

in §

0.

Consider the

following

assertions:

Y(k, N), k

&#x3E;

0, n

&#x3E; 3: there exists a chain Y of

y(k, N)

smooth

conics in PN with

surjective;

if either k &#x3E; 6 or k &#x3E;

2,

N &#x3E;

4,

or N &#x3E;

6,

there is such a Y which is contained in an

integral hyper-

surface of

degree

k.

Z(k, N), k

&#x3E;

0,

N &#x3E; 3: there exists a chain Y of

c(k, N) -~- k -

1

smooth conics in PN with

injective.

W(k,a,N,j), k&#x3E;2, N&#x3E;3,

for every

subset S c PN,

in linear

general position,,

and every

A,

B E

~’, B,

there is a curve Y c PN such that:

(a)

Y n S = -

~A., B~,

is

surjective

and a

general hypersurface

of

degree k containing

is

irreducible;

H(k, N), k

&#x3E;

0,

N &#x3E; 3: there exists a curve

such that:

(11)

(b)

T is a bamboo of

degree 2j(k, N) intersecting Z exactly

at a

point,

say

P,

and

quasi-transversally;

P is a

point

in a final line of

T;

(e) rY,N(k)

is

bijective.

W (k,

a,

N, j )

and

H(k, N)

are

slight

modifications of the assertions of

[4]

with the same name. In

[4] Y(k, N)

and

Z(k, N)

were

proved

for N &#x3E; 4. The same method

(Horace’s

construction

using

a

hyper- plane) gives Y( k, 4), Z(k, 4), using

2.2 if k &#x3E; 30

(plus

a numerical

lemma:

c 2e(k, 4)

- 2 &#x3E;

r(k, 0, 3)/2 + (k2/6) +

10k if k &#x3E; 30 » whose

proof

is left to the

reader), using

2.3 if k 31.

In the same way we

get

the «new» assertions

W(k,

ac,

4, j), H(k, 4).

Then the

proofs

of the « new»

W(k,

a,

N, j), H(k, N),

N &#x3E;

4,

are done

by

induction as in

[4];

the cases with N = 4

simplify

the discussion of the cases with low k for N = 5

given

in

[4],

6.4. Then theorem 1 is

proved

in

3,

in the same way the

corresponding

theorem is

proved

for

PN ,

N &#x3E;

4,

in

[4],

end

of §

7. The same

proof

works for

theorem

0.2, although

a

simpler

one could be done in this case,

adding

in a

hyperplane

irreducible

hyperelliptic

curves.

4. After this paper was

typed,

we read

[6].

It is

elementary

to

show how the results of

[6],

th.

4, 5,

about

sygygies

of

general

canonical

curves can be

adapted

to

give

results about

syzygies

of

general k-gonal

curves for suitable k. We have:

PROPOSITION 4.1. Let X be a

non-hyperlliptic k-gonal

genus n

curves. Assume

g~,2(X)

= 0 for an

integer

p with

1 p n -

3

p c k - 3.

Then

(a)

If C is a

general k-gonal

curve of genus n

+

p

+ 1,

then

(b)

If C is a

general k-gonal

curve of genus m, where

For the

proof

of

4.1,

it is sufficient to take in the

proof

of

[6],

th.

4,

as divisor q,

+

q2

+

...

-~-

qrp+2 a divisor contained in

a gk (strictly

contained

by

the

assumption k

&#x3E; p

+ 2)

and as

smoothing

of X U Y

an admissible k-cover. Then from 4.1 we

get

verbatim the

following

improved

version

of [6],

th. 5:

(12)

PROPOSITION 4.2. Let C be a

general k-gonal

curve of genus g.

(a) K2,2(C)

= 0 if

g &#x3E; 7

and either k &#x3E; 5 or

g -1,

2 mod

(3)

and

k = 5.

(b ) ~3,2(5)

- 0 if

g &#x3E; 9

and either k &#x3E; 6 or 2 mod

(4)

and

k = 6.

(c) KI,2(C)

= 0 if

6, and g =1, 2

mod

(5).

REFERENCES

[1] E. BALLICO - PH. ELLIA, Generic curves

of

small genus in P3 are

of

max-

imal rank, Math. Ann., 264

(1983),

pp. 211-225.

[2] E. BALLICO - PH. ELLIA, On

degeneration of projective

curves, in

Algebraic Geometry-Open

Problems, Proc. Conf., Ravello, 1982,

Springer

Lect.

Notes Math. 997 (1983), pp. 1-15.

[3] E. BALLICO - PH. ELLIA, On the

hypersurfaces containing

a

general

pro-

jective

curve,

Compositio

Math., 60 (1986), pp. 85-95.

[4] E. BALLICO - PH. ELLIA, Postulation

of general

canonical curves in

PN,

N ~ 4, Boll. U.M.I., Sez. D (1986), pp. 103-133.

[5] M.-C. CHANG, Postulation

of

canonical curves in

P3,

Math. Ann., 274

(1986),

pp. 27-30.

[6] L. EIN, A remark on the

syzygies of

the

generic

canonical curves, J. Diff.

Geom., 26 (1987), pp. 361-365.

[7] J. HARRIS - D. MUMFORD, On the Kodaira dimension

of

the moduli space

of

curves, Invent. Math., 67 (1982), pp. 23-86.

[8] A. HIRSCHOWITZ, Sur la

postulation générique

des courbes rationnelles, Acta Math., 146 (1981), pp. 209-230.

[9] A. TANNENBAUM,

Deformation of

space curves, Arch. Math., 34

(1980),

pp. 37-42.

Manoscritto pervenuto in redazione il 16 ottobre 1987.

Références

Documents relatifs

One can also define a projective structure by a holomorphic Riemannian metric, by considering its geodesics defined by Levi-Civita (affine) connection.. But it is not true that

These results are very similar to [15, Theorem 1]; however, a foliated structure (or Riccati foliation) needs not lift as a linear first integral for the affine connection,

The irreducibility of 2^ g then follows from the one of My We shall find it convenient to translate our problem about plane curves into one about g^s on smooth genus ^curves, as we

the period matrix means that the kernel of the isogeny should be a Lagrangian subgroup of the group of points of order 2 in the jacobian.1. In this article we give

the minimal multiple of the minimal class which becomes effective on a (general) principally polarized abelian g-fold are also curves of minimal

Toute utilisa- tion commerciale ou impression systématique est constitutive d’une in- fraction pénale.. Toute copie ou impression de ce fichier doit conte- nir la présente mention

One says that an abelian variety A (resp. Jac C ) is geometrially isomorphi to a produt of supersingular ellipti urves.. However, the desription of the isogeny

We first construct formal foliations on such neighborhoods with holonomy vanishing along many loops, then give the formal / analytic classification of neighborhoods equipped with